• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Your stress testing routine?

B

BababooeyHTJ

Supreme [H]ardness
Joined
Jan 21, 2009
Messages
6,951
I'm starting this thread to see what everyone else does when stress testing an overclock for 24/7 use. Also what chip are you using? How much ram, etc?

I am running an i7 860 and a 2x2GB kit of ram atm. I normally run a couple hours of prime small fft to test for preliminary stability. Then an hour or so run of Large fft with all of my ram enabled. Then about 40-50 runs of Intel Burn Test. Then a couple of hours of blend with all of my ram enabled to use. Then is my overnight small fft run. Last is a quick run of blend and gpu tool to scan for artifacts.

What do you guys do?
 
My routine is nowhere near that intensive. I normally run Prime95 blend for a couple hours, and if that passes I'll run IntelBurnTest overnight to check temps. That is about the extent of it for me.
 
small fft prime for as long as i can. i usually get a 12+ hour window at some point. intel burn test for 50 runs. i normally dont worry too much about stressing the ram, as my normal usage does that for me.

while running prime95 i run a windowed 3d app like furmark or the heaven benchmark and let that loop for a couple hours to find out my temp ceiling for the whole case.
 
I just do Prime Torture Test + 3dmark for several hours, to simulate maximum load. It's what I've been using for years, and it produces very stable overclocks.
 
i just prime95 til the temps stop moving up, then let it run an additional 20 minutes or so. if all is good, i don't find any more reason to keep it going, really
 
I do 10 passes of Intel Burn Test using as much ram as possible and then run Prime 95 blend overnight. I have yet to have that combination fail me.
 
When trying to find my highest overclock, I start out testing with LinX, upping the clocks every 5-10 mins I go without an error.

Once I think I have final clocks set, I start extended testing, which is:

Minimum of 6 hours of P95 on in-place-fft's
Minimum of 6 hours of LinX 0.64 set to max problem size.
Minimum of 6 hours passing MemTest for Windows (Multiple instances to fill up all RAM)

If I pass all of those, I consider the system ready to roll. Anyone who's ever corrupted HDD data due to an unstable overclock can vouge for how important thorough stress testing is.
 
I usually run prime95 for 1-2 hours, super_pi 1mb+32mb a few times, OCCT tests for 10-15 minutes, IBT 10 passes, 3dmark06/vantage 1 run, crysis warhead/gta IV/other games for a few hours. This has always given me rock solid systems (and I typically upgrade components often).

Why people feel it necessary to run stress-tests over night (save for RAM) or for ridiculous amounts of hours -- putting more stress on the CPU/system than could have probably ever been accomplished within their intended use for an entire week or month always makes me laugh: whatever works for them though - not my concern.
 
Rock&Roll said:
Anyone who's ever corrupted HDD data due to an unstable overclock can vouge for how important thorough stress testing is.
Click to expand...

QFT

Just this weekend I failed IBT after 20 or 30 runs. I've passed linpack but failed small fft, I passed small fft, blend, and linpack but failed a custom prime run.
 
how can it corrupt your data? havent heard of this before, thats not a good thing!
 
I usually do 20 passes of LinX. When I get my final clocks, I do 150 passes, then put on P95 blend for 12 hours or so, then CoreDamage to check temps. Then it starts folding. :D
 
OCCT in PS test until I think about stopping.This gets every component.For some reason IBT fails with memory greater than 1gb.
 
On my i7 I found with LinX, not sure about IBT, that it was quicker to show short comings in stability.

On my AM3 boards, I find P95 to be the best. I just hate how long it takes since I'll let it run for 24 Hrs just to make sure. I can pass LinX and fail P95 large ffts after 6 Hrs.

Note: On at least the 1366 i7, there are certain GFlop numbers you should be seeing depending on the speed of the core, uncore, ram, and timings of it all. I found timings to have little impact but they still do effect it slightly. It doesn't matter how much memory you use, the numbers shouldn't vary with increase memory usage. Also important is that you have all non Windows taskbar and background processes not running.

Example: In W7 RC, I would only have TurboV show up in the taskbar outside of the normal Windows icons I have there. Also no other programs running, even Real Temp. So, it is important to know your temperatures are safe before you do this.

Lastly, when you have enough volts and correct balances, your GFlop numbers should vary less than 1GFlop at the very most. Normal for me was .5GFlop variation after 25 loops. It is also important that you use the same LinX version through out all the testing. I've always used 0.6.1 since it never gave me problems unless my overclock was unstable. Its the old antage don't fix was isn't broken.

This process proved reliable for me over and over again in validating the stability of my 920.
 
You must log in or register to reply here.
Back
Top